1024er · jhandsel · May 9, 2022 · May 9, 2022 · May 9, 2022 · May 9, 2022
diff --git a/.gitignore b/.gitignore
@@ -0,0 +1,137 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# Vim
+*.swp
+
+# Repo specific
+cbert_model/
+aug_data_*/
+result/
diff --git a/README.md b/README.md
@@ -1,4 +1,35 @@
-# cbert_aug
+# Changes to make reproducible
+
+Build docker environment:
+```
+docker build -f docker/Dockerfile -t cbert_cuda110 .
+```
+
+Mount `code` directory and connect interactively to instance:
+```
+docker run \
+--gpus all \
+--mount type=bind,source="$(pwd)"/code,target=/cbert/code \
+-it cbert_cuda110 /bin/bash
+```
+
+Fine tune BERT on Subjective/Objective dataset (takes ~20 minutes):
+(see file for other options)
+```
+python3 cbert_finetune.py
+```
+Model is saved to `cbert_model/`.
+
+Generate augmented data for Subjective/Objective dataset:
+```
+python3 cbert_augdata.py --num_train_epochs 1
+```
+The data are output to `aug_data_<SUFFIX>/subj`, where `<SUFFIX>` depends on the
+parameters used for the data augmentation.
+- `train_origin.tsv`: unaugmented data
+- `train.tsv`: unaugmented data + augmented data (appended)
+
+# Original README
 
 Thanks @liuyaxin 's effort to rewrite the code with huggingface's latest transformer library.
 If you want to reproduce the results in paper, you can switch to the develop branch.

diff --git a/args_of_text_classifier.py → code/args_of_text_classifier.py b/args_of_text_classifier.py → code/args_of_text_classifier.py
diff --git a/aug_data/TREC/dev.tsv → code/aug_data/TREC/dev.tsv b/aug_data/TREC/dev.tsv → code/aug_data/TREC/dev.tsv
diff --git a/aug_data/TREC/test.tsv → code/aug_data/TREC/test.tsv b/aug_data/TREC/test.tsv → code/aug_data/TREC/test.tsv
diff --git a/aug_data/TREC/train_origin.tsv → code/aug_data/TREC/train_origin.tsv b/aug_data/TREC/train_origin.tsv → code/aug_data/TREC/train_origin.tsv
diff --git a/aug_data/mpqa/dev.tsv → code/aug_data/mpqa/dev.tsv b/aug_data/mpqa/dev.tsv → code/aug_data/mpqa/dev.tsv
diff --git a/aug_data/mpqa/test.tsv → code/aug_data/mpqa/test.tsv b/aug_data/mpqa/test.tsv → code/aug_data/mpqa/test.tsv
diff --git a/aug_data/mpqa/train_origin.tsv → code/aug_data/mpqa/train_origin.tsv b/aug_data/mpqa/train_origin.tsv → code/aug_data/mpqa/train_origin.tsv
diff --git a/aug_data/rt-polarity/dev.tsv → code/aug_data/rt-polarity/dev.tsv b/aug_data/rt-polarity/dev.tsv → code/aug_data/rt-polarity/dev.tsv
diff --git a/aug_data/rt-polarity/test.tsv → code/aug_data/rt-polarity/test.tsv b/aug_data/rt-polarity/test.tsv → code/aug_data/rt-polarity/test.tsv
diff --git a/aug_data/rt-polarity/train_origin.tsv → code/aug_data/rt-polarity/train_origin.tsv b/aug_data/rt-polarity/train_origin.tsv → code/aug_data/rt-polarity/train_origin.tsv
diff --git a/aug_data/stsa.binary/dev.tsv → code/aug_data/stsa.binary/dev.tsv b/aug_data/stsa.binary/dev.tsv → code/aug_data/stsa.binary/dev.tsv
diff --git a/aug_data/stsa.binary/test.tsv → code/aug_data/stsa.binary/test.tsv b/aug_data/stsa.binary/test.tsv → code/aug_data/stsa.binary/test.tsv
diff --git a/aug_data/stsa.binary/train_origin.tsv → code/aug_data/stsa.binary/train_origin.tsv b/aug_data/stsa.binary/train_origin.tsv → code/aug_data/stsa.binary/train_origin.tsv
diff --git a/aug_data/stsa.fine/dev.tsv → code/aug_data/stsa.fine/dev.tsv b/aug_data/stsa.fine/dev.tsv → code/aug_data/stsa.fine/dev.tsv
diff --git a/aug_data/stsa.fine/test.tsv → code/aug_data/stsa.fine/test.tsv b/aug_data/stsa.fine/test.tsv → code/aug_data/stsa.fine/test.tsv
diff --git a/aug_data/stsa.fine/train_origin.tsv → code/aug_data/stsa.fine/train_origin.tsv b/aug_data/stsa.fine/train_origin.tsv → code/aug_data/stsa.fine/train_origin.tsv
diff --git a/aug_data/subj/dev.tsv → code/aug_data/subj/dev.tsv b/aug_data/subj/dev.tsv → code/aug_data/subj/dev.tsv
diff --git a/aug_data/subj/test.tsv → code/aug_data/subj/test.tsv b/aug_data/subj/test.tsv → code/aug_data/subj/test.tsv
diff --git a/aug_data/subj/train_origin.tsv → code/aug_data/subj/train_origin.tsv b/aug_data/subj/train_origin.tsv → code/aug_data/subj/train_origin.tsv
diff --git a/cbert_augdata.py → code/cbert_augdata.py b/cbert_augdata.py → code/cbert_augdata.py
diff --git a/cbert_finetune.py → code/cbert_finetune.py b/cbert_finetune.py → code/cbert_finetune.py
diff --git a/cbert_utils.py → code/cbert_utils.py b/cbert_utils.py → code/cbert_utils.py
diff --git a/datasets/TREC/dev.tsv → code/datasets/TREC/dev.tsv b/datasets/TREC/dev.tsv → code/datasets/TREC/dev.tsv
diff --git a/datasets/TREC/test.tsv → code/datasets/TREC/test.tsv b/datasets/TREC/test.tsv → code/datasets/TREC/test.tsv
diff --git a/datasets/TREC/train.tsv → code/datasets/TREC/train.tsv b/datasets/TREC/train.tsv → code/datasets/TREC/train.tsv
diff --git a/datasets/TREC/train_dev.tsv → code/datasets/TREC/train_dev.tsv b/datasets/TREC/train_dev.tsv → code/datasets/TREC/train_dev.tsv
diff --git a/datasets/mpqa/dev.tsv → code/datasets/mpqa/dev.tsv b/datasets/mpqa/dev.tsv → code/datasets/mpqa/dev.tsv
diff --git a/datasets/mpqa/test.tsv → code/datasets/mpqa/test.tsv b/datasets/mpqa/test.tsv → code/datasets/mpqa/test.tsv
diff --git a/datasets/mpqa/train.tsv → code/datasets/mpqa/train.tsv b/datasets/mpqa/train.tsv → code/datasets/mpqa/train.tsv
diff --git a/datasets/mpqa/train_dev.tsv → code/datasets/mpqa/train_dev.tsv b/datasets/mpqa/train_dev.tsv → code/datasets/mpqa/train_dev.tsv
diff --git a/datasets/rt-polarity/dev.tsv → code/datasets/rt-polarity/dev.tsv b/datasets/rt-polarity/dev.tsv → code/datasets/rt-polarity/dev.tsv
diff --git a/datasets/rt-polarity/test.tsv → code/datasets/rt-polarity/test.tsv b/datasets/rt-polarity/test.tsv → code/datasets/rt-polarity/test.tsv
diff --git a/datasets/rt-polarity/train.tsv → code/datasets/rt-polarity/train.tsv b/datasets/rt-polarity/train.tsv → code/datasets/rt-polarity/train.tsv
diff --git a/datasets/rt-polarity/train_dev.tsv → code/datasets/rt-polarity/train_dev.tsv b/datasets/rt-polarity/train_dev.tsv → code/datasets/rt-polarity/train_dev.tsv
diff --git a/datasets/stsa.binary/dev.tsv → code/datasets/stsa.binary/dev.tsv b/datasets/stsa.binary/dev.tsv → code/datasets/stsa.binary/dev.tsv
diff --git a/datasets/stsa.binary/test.tsv → code/datasets/stsa.binary/test.tsv b/datasets/stsa.binary/test.tsv → code/datasets/stsa.binary/test.tsv
diff --git a/datasets/stsa.binary/train.tsv → code/datasets/stsa.binary/train.tsv b/datasets/stsa.binary/train.tsv → code/datasets/stsa.binary/train.tsv
diff --git a/datasets/stsa.binary/train_dev.tsv → code/datasets/stsa.binary/train_dev.tsv b/datasets/stsa.binary/train_dev.tsv → code/datasets/stsa.binary/train_dev.tsv
diff --git a/datasets/stsa.fine/dev.tsv → code/datasets/stsa.fine/dev.tsv b/datasets/stsa.fine/dev.tsv → code/datasets/stsa.fine/dev.tsv
diff --git a/datasets/stsa.fine/test.tsv → code/datasets/stsa.fine/test.tsv b/datasets/stsa.fine/test.tsv → code/datasets/stsa.fine/test.tsv
diff --git a/datasets/stsa.fine/train.tsv → code/datasets/stsa.fine/train.tsv b/datasets/stsa.fine/train.tsv → code/datasets/stsa.fine/train.tsv
diff --git a/datasets/stsa.fine/train_dev.tsv → code/datasets/stsa.fine/train_dev.tsv b/datasets/stsa.fine/train_dev.tsv → code/datasets/stsa.fine/train_dev.tsv
diff --git a/datasets/subj/dev.tsv → code/datasets/subj/dev.tsv b/datasets/subj/dev.tsv → code/datasets/subj/dev.tsv
diff --git a/datasets/subj/test.tsv → code/datasets/subj/test.tsv b/datasets/subj/test.tsv → code/datasets/subj/test.tsv
diff --git a/datasets/subj/train.tsv → code/datasets/subj/train.tsv b/datasets/subj/train.tsv → code/datasets/subj/train.tsv
diff --git a/datasets/subj/train_dev.tsv → code/datasets/subj/train_dev.tsv b/datasets/subj/train_dev.tsv → code/datasets/subj/train_dev.tsv
diff --git a/evaluator.py → code/evaluator.py b/evaluator.py → code/evaluator.py
diff --git a/global.config → code/global.config b/global.config → code/global.config
diff --git a/nets.py → code/nets.py b/nets.py → code/nets.py
diff --git a/text_classification/__init__.py → code/text_classification/__init__.py b/text_classification/__init__.py → code/text_classification/__init__.py
diff --git a/text_classification/nets.py → code/text_classification/nets.py b/text_classification/nets.py → code/text_classification/nets.py
diff --git a/text_classification/nlp_utils.py → code/text_classification/nlp_utils.py b/text_classification/nlp_utils.py → code/text_classification/nlp_utils.py
diff --git a/text_classification/text_datasets.py → code/text_classification/text_datasets.py b/text_classification/text_datasets.py → code/text_classification/text_datasets.py
diff --git a/train_text_classifier.py → code/train_text_classifier.py b/train_text_classifier.py → code/train_text_classifier.py
diff --git a/triggers.py → code/triggers.py b/triggers.py → code/triggers.py
diff --git a/utils.py → code/utils.py b/utils.py → code/utils.py
diff --git a/docker/Dockerfile b/docker/Dockerfile
@@ -0,0 +1,24 @@
+from nvidia/cuda:11.0.3-cudnn8-devel-ubuntu20.04
+
+WORKDIR /cbert
+
+COPY requirements.txt requirements.txt
+
+# Disable nvidia repos as their GPG key isn't recognized
+RUN mv -f /etc/apt/sources.list.d/cuda.list /etc/apt/sources.list.d/cuda.list.bak 2>/dev/null; true
+RUN mv -f /etc/apt/sources.list.d/nvidia-ml.list /etc/apt/sources.list.d/nvidia-ml.list.bak 2>/dev/null; true
+
+RUN apt-get update &&\
+    apt-get install --no-install-recommends --no-install-suggests -y gnupg2 \
+                                                                     ca-certificates \
+                                                                     git \
+                                                                     build-essential \
+                                                                     python3-dev \
+                                                                     python3-pip \
+                                                                     wget \
+								     curl \
+                                                                     rsync
+
+
+RUN pip install -r requirements.txt
+WORKDIR /cbert/code
diff --git a/docker/nvidia_gpgkey b/docker/nvidia_gpgkey
@@ -0,0 +1,52 @@
+-----BEGIN PGP PUBLIC KEY BLOCK-----
+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+=9QWY
+-----END PGP PUBLIC KEY BLOCK-----
diff --git a/requirements.txt b/requirements.txt
@@ -0,0 +1,7 @@
+torch==1.11.0
+transformers==2.1.0
+cython==0.29.28
+chainer==7.8.1
+cupy-cuda110==7.8.0
+scikit_learn==1.0.2
+progressbar==2.5
diff --git a/text_classification/__pycache__/__init__.cpython-35.pyc b/text_classification/__pycache__/__init__.cpython-35.pyc
diff --git a/text_classification/__pycache__/__init__.cpython-36.pyc b/text_classification/__pycache__/__init__.cpython-36.pyc
diff --git a/text_classification/__pycache__/__init__.cpython-37.pyc b/text_classification/__pycache__/__init__.cpython-37.pyc
diff --git a/text_classification/__pycache__/nets.cpython-35.pyc b/text_classification/__pycache__/nets.cpython-35.pyc
diff --git a/text_classification/__pycache__/nets.cpython-36.pyc b/text_classification/__pycache__/nets.cpython-36.pyc
diff --git a/text_classification/__pycache__/nets.cpython-37.pyc b/text_classification/__pycache__/nets.cpython-37.pyc
diff --git a/text_classification/__pycache__/nlp_utils.cpython-35.pyc b/text_classification/__pycache__/nlp_utils.cpython-35.pyc
diff --git a/text_classification/__pycache__/nlp_utils.cpython-36.pyc b/text_classification/__pycache__/nlp_utils.cpython-36.pyc
diff --git a/text_classification/__pycache__/nlp_utils.cpython-37.pyc b/text_classification/__pycache__/nlp_utils.cpython-37.pyc
diff --git a/text_classification/__pycache__/text_datasets.cpython-35.pyc b/text_classification/__pycache__/text_datasets.cpython-35.pyc
diff --git a/text_classification/__pycache__/text_datasets.cpython-36.pyc b/text_classification/__pycache__/text_datasets.cpython-36.pyc
diff --git a/text_classification/__pycache__/text_datasets.cpython-37.pyc b/text_classification/__pycache__/text_datasets.cpython-37.pyc